Journal of Experimental Psychology:
Learning, Memory, and Cognition
2003, Vol. 29, No. 4, 524 –531
Copyright 2003 by the American Psychological Association, Inc.
0278-7393/03/$12.00 DOI: 10.1037/0278-7393.29.4.524
Intentional Forgetting Can Increase, Not Decrease, Residual Influences of
To-Be-Forgotten Information
Elizabeth Ligon Bjork and Robert A. Bjork
University of California, Los Angeles
Intentionally forgotten information remains in memory at essentially full strength, as measured by
recognition and priming, but access to that information is impaired, as measured by recall. Given that
pattern, it seemed plausible that intentionally forgotten information might have a greater impact on
certain subsequent judgments than would intentionally remembered information. In 2 experiments,
participants cued to forget nonfamous names were subsequently more likely to make false attributions of
fame to those names than were participants instructed to remember them. These findings implicate
retrieval inhibition as a potent factor in the interplay of recollection and priming in memory and
judgment. They also point to possible unintended consequences of instructions to forget, suppress, or
disregard in legal or social settings.
full strength, even when access to that information, as measured by
recall tests, is impaired (e.g., Basden, Basden, & Gargano, 1993;
E. L. Bjork & Bjork, 1996; E. L. Bjork, Bjork, & Kirkpatrick,
1990). In a study by E. L. Bjork and Bjork (1996), for example,
TBF items primed subsequent completions on a word-fragmentcompletion test to the same extent as did comparable TBR items,
even when retrieval access to the TBF items remained inhibited.
Forgetting—although typically a major complaint people have
about their memories—is often exactly what individuals need to do
to keep their memories functioning effectively in a changing
world. Put differently, as we and others have argued (e.g., E. L.
Bjork & Bjork, 1988, 1996; R. A. Bjork, 1972; Geiselman, Bjork,
& Fishman, 1983; MacLeod, 1998), an efficient memory system
needs some means to suppress information that is out of date and
a possible source of errors and interference.
Research on directed, or intentional, forgetting suggests that a
key mechanism of forgetting, as an updating process, is retrieval
inhibition. When people are cued, explicitly or implicitly, that
previously learned information needs to be forgotten and replaced
with new information, a process is initiated that inhibits the subsequent retrieval of the to-be-forgotten (TBF) information. Such
retrieval inhibition serves to reduce, and can eliminate, the interference due to the TBF information on the subsequent retrieval of
to-be-remembered (TBR) information (e.g., R. A. Bjork, 1972;
E. L. Bjork & Bjork, 1996). As indexed by measures other than
retrieval, however, the TBF information remains in memory, typically at full strength. For example, TBF items can be recognized
as well as comparable TBR items (Block, 1971; Elmes, Adams, &
Roediger, 1970; Geiselman et al., 1983; Gross, Barresi, & Smith,
1970) and can be relearned as readily as TBR items (e.g., Geiselman & Bagheri, 1985; Reed, 1970). Additionally, under certain
conditions, the proactive interference of TBF items that is diminished or eliminated in response to the forget cue can be reinstated
at full strength, such as when TBF items are encountered as foils
on a recognition test for TBR items (E. L. Bjork & Bjork, 1996;
E. L. Bjork, Bjork, & Glenberg, 1973).
Indirect measures of retention also provide evidence that intentionally forgotten information remains in memory at essentially
Possible Consequences of Preserved Storage and
Impaired Access
Given the evidence that intentionally forgotten information continues to reside in memory and to have indirect influences on
behavior, it seems possible that the indirect influences of forgotten
information might actually be greater than those of intentionally
remembered information in certain situations. Of particular concern in this regard would be situations in which indirect influences
could contaminate or impair some judgment; that is, situations in
which optimal performance depends on recollection to check or
mitigate undesirable indirect effects. To create such a situation in
which to test this possibility, we capitalized on the false-fame
paradigm and results of Jacoby and his colleagues (e.g., Jacoby,
Kelley, Brown, & Jasechko, 1989).
In the false-fame paradigm of Jacoby et al. (1989), participants
were first asked to read a list of names aloud as part of a pronunciation task. Following this task, some participants were immediately required to engage in a fame-judgment task, whereas others
did so on the following day. In this task, participants saw a series
of famous and nonfamous names and, for each, were asked to
judge whether it was the name of a famous person. Half of the
presented nonfamous names had appeared in the earlier pronunciation task and thus were old nonfamous names, and half were new.
The remaining names were those of moderately famous people,
that is, names recognizable as famous but not so famous that the
reason for their fame could generally be specified. Before beginning this task, participants were told that some of the names they
would see had been presented in the earlier pronunciation task, and
they were reminded that all of the names in that task had been
Elizabeth Ligon Bjork and Robert A. Bjork, Department of Psychology,
University of California, Los Angeles.
Correspondence concerning this article should be addressed to Elizabeth
Ligon Bjork, Department of Psychology, University of California, Los
Angeles, California 90095-1563. E-mail: elbjork@psych.ucla.edu
524
RESIDUAL INFLUENCES OF “FORGOTTEN” INFORMATION
nonfamous. Thus, if they recognized a name during the famejudgment task as one from the earlier task, they could know it was
a nonfamous name. In essence, then, participants were given a rule
that could be used whenever they encountered a name that seemed
familiar but that they could not specifically recollect as being the
name of a famous person. If they could recollect that it had
appeared in the pronunciation task, then they could know to judge
it as being nonfamous despite its familiarity.
Of primary interest was whether participants in the immediate
and delayed conditions would respond differently to old and new
nonfamous names, and their pattern of responding to such names
proved to be markedly different. On the immediate test, the proportion of old nonfamous names judged to be famous was .12, and
the corresponding proportion for new nonfamous names was .21.
After 24 hrs, however, the proportion of old nonfamous names
called famous was .16, whereas the corresponding proportion for
new nonfamous names was .08.
As interpreted by Jacoby et al. (1989), this complete reversal
in the pattern of false-fame judgments arises because the priming effects attributable to a nonfamous name’s prior occurrence
in the pronunciation task persist across 24 hrs, whereas the
ability to recollect that a given name appeared in that task is
largely lost at that point. The combination of preserved priming,
making the name seem familiar, and impaired recollection thus
creates an attribution problem. To what should that familiarity
be attributed? For participants tested immediately, it can typically be attributed to the name’s occurrence in the earlier task,
but if the ability to recall whether a given name occurred during
that earlier task is impaired, as was apparently the case at 24
hrs, an attribution problem arises. One may think that a given
name, “Sebastian Weisdorf,” for example, is familiar because it
is the name of an artist or actor to whom one has been exposed
in the media. To the extent a participant arrives at that decision,
Sebastian Weisdorf, to quote Jacoby et al. (1989), can become
“famous overnight” (p. 326).
The Present Paradigm and Hypothesis
In the false-fame paradigm, whether a given nonfamous name is
recollected as having occurred earlier has a dramatic effect on later
judgments of fame. If recollection succeeds, the name is surely not
famous, but if recollection fails, the familiarity of the name might
be attributable to its occurrence in the media rather than in the
present context. Thus, this paradigm represents just the type of
situation in which we suspected information intentionally forgotten might have a greater effect on one’s judgment than would
information intentionally remembered.
To test this possibility, we combined features of the falsefame paradigm and the list method of directed forgetting into a
single paradigm that included the three tasks illustrated in
Figure 1. During the memory task, all participants received a
first list of nonfamous names to study for a later memory test.
Participants receiving a forget–remember (FR) list were then
told that this first list had been presented in error, that they had
accidentally been shown a list of nonfamous names, whereas
they should have been shown a list of famous names. They were
thus instructed to forget that first list of names, and a second list
of famous names was presented. Participants receiving a
remember–remember (RR) list were told that the first list con-
525
Figure 1. Pictorial illustration of the paradigm and tasks used in Experiment 1. RR ⫽ remember–remember condition; FR ⫽ forget–remember
condition.
tained only nonfamous names and they would now be presented
with a second list to remember that would contain only famous
names. Next, all participants received a cued-recall test for a
few famous names to give a sense of closure to the memory
task. Then, following a brief questionnaire task, all participants
engaged in a fame-judgment task in which some of the famous
and nonfamous names presented were ones that had also been
presented earlier in the memory task.
Our primary interest was how old nonfamous names would be
judged when participants were cued to forget them versus to
remember them. We hypothesized that if an instruction to forget
the list of nonfamous names does not diminish the indirect effects
of a given nonfamous name’s prior presentation but does impair
participants’ ability to recollect that the name was on that list, there
should be an increased tendency for participants to label such
names incorrectly as famous. In short, the tendency for participants
to misattribute the familiarity of old nonfamous names to realworld fame should be exacerbated when they are instructed to
forget those names rather than to remember them.
BJORK AND BJORK
526
Experiment 1
Method
Participants. A total of 48 students attending the University of California, Los Angeles, participated in partial fulfillment of an introductory
psychology course requirement.
Design. The design was a four-way mixed factorial. The one betweensubjects variable, list type (FR vs. RR), had 24 participants assigned to
each condition. The remaining within-subjects variables were fame status
(famous vs. nonfamous), presentation status (old vs. new), and judged
order (first half vs. second half) of the judged names. The dependent
variable was the proportion of famous judgments made to each type of
name during the fame-judgment task.
Materials. Two memory lists (Versions 1 and 2), each containing 24
nonfamous names in the precue part of the list and 24 famous names in the
postcue part of the list, were constructed from a subset of the names used
by Jacoby et al. (1989), and half the participants (n ⫽ 12) in each memory
condition were randomly assigned to receive one of these versions, which
were created as follows. From an original pool of 36 nonfamous names, 24
names were randomly selected to constitute the precue part of Version 1.
From each one third of this list, 4 names were randomly selected to appear
as the old nonfamous names in the fame-judgment task, and these 12 names
were randomly replaced by the remaining 12 nonfamous names from the
original pool to create the precue part of Version 2. Thus, the precue parts
of Versions 1 and 2 contained 12 identical and 12 different nonfamous
names. For participants receiving Version 1, the 12 different names in
Version 1 appeared as old nonfamous names and the 12 different names in
Version 2 appeared as new nonfamous names in the fame-judgment task.
For participants receiving Version 2, the reverse was true. The two postcue
parts of each list, which contained only famous names, were constructed in
the identical manner and then randomly assigned to be the postcue part of
Version 1 or Version 2.
Owing to this construction and the assignment of participants to the two
versions, names appearing as old nonfamous and famous names, respectively, in the fame-judgment task for a random half of the participants
appeared as new nonfamous and famous names, respectively, for the other
half of the participants and vice versa. Thus, across participants and for
both memory conditions, the same names appeared equally often as old and
new nonfamous names and as old and new famous names, respectively, in
the fame-judgment task.
A questionnaire consisting of geography and history questions was
constructed to serve as an engaging 5-min distractor task between the
memory and fame-judgment tasks.
Four versions of the fame-judgment list were created. Versions 1 and 2
corresponded to Versions 1 and 2 of the memory list in terms of the
counterbalancing of old and new names. Thus, items appearing as old
nonfamous names and old famous names in Version 1 appeared as new
nonfamous names and new famous names, respectively, in Version 2. Each
version presented 48 names, 12 of each type, randomly ordered except that
the same type could not appear more than twice in succession and 6 of each
type had to occur in the first and second halves so that the second half
would be a replication of the first. Versions 3 and 4 were created by
reversing the first and second halves of Versions 1 and 2, respectively. For
participants receiving Version 1 and 2 of the memory list, half (n ⫽
12)—in both memory conditions—were randomly assigned to receive
Versions 1 and 3 or Versions 2 and 4, respectively, of the fame-judgment
list. This construction and assignment of participants allowed a comparison
of performance, unbiased by specific item effects, across the two halves of
the fame-judgment task.
Procedure. Each participant was seated at a computer in a testing room
and informed that he or she would be performing several tasks to see if a
relationship existed between one’s ability to learn names and to perform
other tasks. All participants were told that the first task would present a list
of names on the screen, at a rate of 5 s each, and they were to try to learn
them for a later memory test. The experimenter explained that after
beginning the program, he would leave the room and that soon thereafter
a message would appear instructing the participant to start the list when
ready by hitting any key. The experimenter then left the room, and the
participant began the list when ready.
At the end of the list, the experimenter returned to the testing room. At
this point, for FR participants only, he pretended to discover that he had
made a mistake, explaining that he had presented a list of nonfamous
names when he should have presented a list of famous names. The
participant was then told to forget all the nonfamous names that had been
presented by mistake, that they would not be on the memory test, and that
the participant should now try to learn the correct list of famous names to
be presented next. The experimenter further clarified that the famous
names would not be those of very famous people, only slightly famous
people, so the participant should not be disconcerted if he or she was not
able to recognize all of them. As before, the experimenter then started the
program and left the room. When ready, the participant began the list of
famous names, which also appeared at a 5-s rate.
For RR participants, the experimenter, upon returning to the testing
room, explained that the names just presented had all been nonfamous ones
and that the participant should try to keep remembering them for a later
memory test. He then explained that a list of famous names would be
presented next and that the participant should try to learn these names as
well for the later test. From this point, instructions concerning the list of
famous names were the same as in the FR condition.
After the last famous name was presented, the experimenter returned and
explained that a cued-recall test for some of the names would now be
given. In this test, the first name and some letters of the last name appeared
as cues for six of the famous names, and participants typed their responses
into the keyboard. The purpose of the test was to give a sense of closure to
the memory task, no responses were collected, and no name tested appeared in the fame-judgment task later.
After the cued-recall test, participants were given 5 min to work on the
questionnaire, after which the experimenter returned to the testing room. At
that point, the experimenter informed participants that the next task would
assess their knowledge of famous people. They were told that names would
be presented, one at a time, and that they were to decide whether each name
was that of a famous person and then press one of two keys on the number
pad of the keyboard labeled y (for yes) and n (for no), doing so as quickly
and accurately as possible. They were also informed that the famous names
presented would be those of only slightly famous people and that they
would not be asked to report the basis for their decision.
The above instructions were delivered to all participants. Thereafter, for
FR participants, the experimenter went on to say, in an ad-lib fashion, that
some of the nonfamous names appearing in the fame-judgment task might
be ones that had been in the list shown to them earlier by mistake. They
were then reminded that this incorrect list had presented only nonfamous
names. Thus, it was pointed out, if they were trying to decide whether a
given name was famous, and just happened to remember that it was a name
on that first list presented by mistake, then they could be sure that it was
not a famous name. They were also informed that some of the famous
names appearing in the fame-judgment task might be ones that were
presented in the second list—that is, the correct list. Thus, again, if they
were trying to decide whether a given name was famous and could
remember that it was a name presented on the second list, then they could
be sure that it was the name of a famous person. For RR participants, the
experimenter, also in an ad-lib manner, gave the same advice but with
reference to the “first list” and “second list,” rather than the “list presented
by mistake” and the “correct” list.
Results
The effect of judged order, included in Experiment 1 as a way
of checking whether performance changed over the course of the
RESIDUAL INFLUENCES OF “FORGOTTEN” INFORMATION
fame-judgment task, was not significant by a planned comparison,
F(1, 40) ⫽ 0.27, MSE ⫽ 0.97, p ⬎ .50, so the results presented
below have been collapsed across the first and second halves of the
judgment task.
Famous names. The proportions of old and new famous names
judged to be famous by FR and RR participants are shown in
Figure 2A. In this and all other comparisons between old and new
names, it should be remembered that across participants, given the
counterbalancing procedures, nothing is different about the two
sets of names other than whether they were or were not presented
in the experiment. As appears to be the case in Figure 2, and as
confirmed by planned comparisons, the proportions of famous
judgments made to old famous names by FR and RR participants
did not differ significantly (Ms ⫽ .82 and .84, respectively), F(1,
40) ⫽ 0.13, MSE ⫽ 1.26, p ⬎ .50; nor did the proportions of
famous judgments made to new famous names by FR and RR
participants (Ms ⫽ .42 and .51, respectively), F(1, 40) ⫽ 1.80,
MSE ⫽ 3.34, p ⫽ .19.
Nonfamous names. The proportions of old and new nonfamous names judged to be famous (i.e., to which participants made
false judgments of fame) are presented in Figure 2B for FR and RR
participants. In terms of the present hypothesis, the most important
result in Figure 2B is the apparent greater proportion of famous
judgments made to old nonfamous names by FR participants (M ⫽
.52) than by RR participants (M ⫽ .36). And, indeed, this difference was revealed to be significant by a planned comparison, F(1,
40) ⫽ 7.37, MSE ⫽ 2.99, p ⫽ .01. In contrast, the proportion of
famous judgments made to new nonfamous names by FR participants (M ⫽ .21) did not differ from that made by RR participants
(M ⫽ .19), F(1, 40) ⫽ 0.19, MSE ⫽ 1.96, p ⬎ .50.
Thus, as predicted, the FR participants, who had been instructed
to forget the nonfamous names presented in the memory task,
made false judgments of fame significantly more often to such
names than did the RR participants, who had been instructed to
keep remembering them. Also of note in Figure 2 is the difference
in the proportions of famous judgments made by the FR participants to new famous names (see Figure 2A) versus old nonfamous
527
names (see Figure 2B). Although, unlike the former comparisons,
this one involves different sets of names, it is interesting that FR
participants judged more old nonfamous names that they had been
instructed to forget to be the names of actual famous people
(which, of course, they were not) than they judged new famous
names to be the names of actual famous people (which, of course,
they were). In contrast, the judgments of RR participants were
more veridical: RR participants were more likely to judge a new
famous name to be that of a famous person than they were to judge
an old nonfamous name as such.
Discussion
One possible explanation of the observed vulnerability of FR
participants for making false judgments of fame would be in terms
of a problem of source memory. As discussed earlier, research
shows that indirect memory effects of intentionally forgotten information persist and that recognition memory for TBF items (at
least as measured by yes–no and forced-choice tests) is typically
not impaired compared with that for TBR items. Hence, during the
present fame-judgment task, one might expect (a) old nonfamous
names to seem more familiar than new nonfamous names to both
FR and RR participants and (b) that FR and RR participants might
be equally capable of recognizing that an old nonfamous name had
been presented earlier in the memory task. Thus, one might also
expect FR and RR participants to be equally capable of correctly
interpreting the familiarity evoked by such names as arising from
their earlier occurrence in the experiment, rather than from realworld fame.
In the present paradigm, however, for the familiarity evoked by
old nonfamous names to be interpreted correctly, participants have
to be able to do more than just recognize that the name had been
presented previously. Beyond simply recognizing that an old nonfamous name occurred earlier in the experiment, they must be able
to recollect where in the memory task it occurred. Specifically,
they must recollect that it had been presented in the precue part of
the memory list— exactly what FR participants, given their inhib-
Figure 2. Proportions of old and new famous names judged famous (A) and proportions of old and new
nonfamous names judged famous (B) in Experiment 1. RR ⫽ remember–remember list; FR ⫽ forget–remember
list.
BJORK AND BJORK
528
ited retrieval access to the precue list, would have trouble doing
relative to RR participants. Thus, instructing FR participants to
forget the precue part of the memory list could have created a
serious source-memory problem for them relative to the RR
participants.
We hypothesized that if this source-memory problem was primarily of a list-discrimination nature, then FR participants might
frequently be led to interpret the familiarity evoked by an old
nonfamous name as indicating that it was a name presented in the
postcue part of the list, which they knew to contain only famous
names, and thus be led to judge it incorrectly as being famous. In
contrast, we believed that RR participants would be better at
recollecting the specific list membership of an old nonfamous
name that they recognized and could use such information to keep
from making false judgments of fame. That FR participants had
this list-discrimination type of source-memory problem would, in
our view, also be consistent with the previous observation of
Geiselman et al. (1983) that FR participants tended to judge the
TBF items that they were able to recall as belonging incorrectly to
the second (or postcue) list rather than to the first (or precue) list
in which they had actually been presented.
It is also possible, however, that the source-memory problem
leading FR participants to make more misattributions of real-world
fame to old nonfamous names is not limited to that of a list
discrimination problem, but is of a more severe or extensive
nature. Perhaps, upon encountering an old nonfamous name, although it often evokes a feeling of familiarity, FR participants are
frequently not able to recollect that it had been presented earlier in
the experiment at all. Although we knew from past results that
recognition performance for TBF items is usually not impaired as
compared with that for TBR items in standard recognition-memory
tasks, we could not necessarily assume that this would be the case
for the present task. Thus, we conducted Experiment 2 to try to
distinguish between these two possible types of source-memory
problems.
Experiment 2
In Experiment 2, we repeated the design of Experiment 1 but
required an additional response during the fame-judgment task.
After each judgment of fame, we asked participants to indicate
whether they could remember if that name happened to be one that
had occurred earlier in the memory task, with the idea that the
resulting joint responses—famous–yes, famous–no, nonfamous–
yes, nonfamous–no—would allow us to determine more precisely
the nature of the source-memory problem suffered by the FR
participants of Experiment 1.
To illustrate, suppose their source-memory problem had been of
a list-discrimination nature. If so, then the FR participants in
Experiment 2 should mainly respond yes following a judgment of
fame to an old nonfamous name. Alternatively, suppose their
source-memory problem had been of a more extensive nature than
not being able to recollect the list membership of an old nonfamous name that seemed familiar. Suppose, instead, they were
frequently not able to recognize that the name had been presented
earlier at all. If so, then the FR participants in Experiment 2 should
predominantly respond no following a judgment of fame to an old
nonfamous name.
Method
As in Experiment 1, 48 students attending the University of California,
Los Angeles, participated, with 24 individuals assigned randomly to each
memory condition. The design was identical to that of Experiment 1,
except for the second judgment made during the fame-judgment task. The
materials and procedure were also identical, except for the additional
instructions to all participants that after making their judgment of fame to
a given name they then needed to say whether they thought that name was
one that had been presented earlier in the memory task. The question:
Presented today? appeared on the screen after each name was judged as
famous or not famous, and participants were asked to respond by pressing
a yes or no key.
Results and Discussion
The effect of judged order was found to be nonsignificant, F(1,
40) ⫽ 1.01, MSE ⫽ 1.04, p ⫽ .32; in the results below, therefore,
the data have been averaged across the first and second halves of
the fame-judgment task.
Fame judgments. The proportions of old and new famous
names judged to be famous by FR and RR participants are shown
in Figure 3A. As appears to be the case, and as confirmed by
planned comparisons, the proportions of famous judgments made
to old famous names by FR participants (M ⫽ .75) and RR
participants (M ⫽ .75) and to new famous names by FR participants (M ⫽ .40) and RR participants (M ⫽ .42) did not differ, F(1,
40) ⫽ 0.38, MSE ⫽ 0.51, p ⬎ .50; F(1, 40) ⫽ 2.00, MSE ⫽ 4.17,
p ⫽ .16, respectively.
The proportions of old and new nonfamous names judged as
famous (i.e., to which false judgments of fame were made) are
presented for FR and RR participants in Figure 3B. As before, the
result of most interest is the greater proportion of famous judgments made to old nonfamous names by FR participants (M ⫽ .46)
as compared with RR participants (M ⫽ .27), a difference revealed
to be significant by a planned comparison, F(1, 40) ⫽ 11.66,
MSE ⫽ 23.01, p ⬍ .01. In contrast, the proportions of famous
judgments made to new nonfamous names by FR participants
(M ⫽ .21) and RR participants (M ⫽ .19) did not differ, F(1,
40) ⫽ 1.86, MSE ⫽ 2.67, p ⫽ .18. Thus, the pattern of results
obtained in Experiment 2 constitutes a striking replication of that
observed in Experiment 1. The critical difference—judgments of
fame made to old nonfamous names by FR and RR participants—
was again highly significant, with participants who had been told
to forget the nonfamous names making significantly more false
judgments of fame to such names than were participants instructed
to remember them.
Joint responses. The top panel of Table 1 presents the proportions of nonfamous–yes, famous–yes, and famous–no joint responses made to old nonfamous names by FR and RR participants,
and next to each response, we have indicated what, in our view,
that type of response reflects. The joint response of nonfamous–
yes, for example, reflects accurate remembering because participants would report both (a) that an old nonfamous name was not
the name of a famous person and (b) that it had been presented
earlier only if they could correctly remember it as having been
presented in the precue part of the memory list (which they knew
contained only nonfamous names). The other two joint responses,
however, can each represent a type of source-memory problem.
Famous–yes can be thought to reflect failed list discrimination
RESIDUAL INFLUENCES OF “FORGOTTEN” INFORMATION
529
Figure 3. Proportions of old and new famous names judged famous (A) and proportions of old and new
nonfamous names judged famous (B) in Experiment 2. RR ⫽ remember–remember list; FR ⫽ forget–remember
list.
because participants would incorrectly report both (a) that an old
nonfamous name was the name of a famous person and (b) that it
had occurred earlier in the memory task, only if they incorrectly
remembered it as having appeared in the postcue list (which they
knew contained only famous names). Famous–no, on the other
hand, reflects episodic forgetting because participants would incorrectly report both (a) that an old nonfamous name was the name
of a famous person and (b) that it had not been presented earlier,
only if they were not able to recollect it having occurred at all
during the memory task.
For all of the joint responses shown in the top panel of Table 1,
planned comparisons revealed significant differences in the proportions of times each was made by FR versus RR participants,
F(1, 40) ⫽ 14.16, MSE ⫽ 10,072.48, p ⬍ .01; F(1, 40) ⫽ 4.05,
MSE ⫽ 1,807.96, p ⫽ .05; F(1, 40) ⫽ 11.00, MSE ⫽ 1,400.64,
p ⬍ .01, for the responses of nonfamous–yes, famous–yes, and
famous–no, respectively. These results thus indicate a somewhat
Table 1
Joint Response Proportions Made to Old (Precue) Nonfamous
and Old (Postcue) Famous Names in the FR and RR Conditions
of Experiment 2
Condition
Joint response
What Joint
response reflects
FR
RR
.36
.29
.12
.57
.20
.04
Old nonfamous names
Nonfamous–yes
Famous–yes
Famous–no
Accurate remembering
Failed list discrimination
Episodic forgetting
Old famous names
Famous–yes
Nonfamous–yes
Famous–no
Accurate remembering
Failed list discrimination
Episodic forgetting
.73
.14
.05
Note. FR ⫽ forget–remember list; RR ⫽ remember–remember list.
.72
.14
.08
surprising conclusion: It is not one type of source-memory problem versus the other from which FR participants are suffering.
Rather, the instruction to forget appears to have created both types
of source-memory problems. Thus, unlike previous studies indicating that TBF items can be recognized as well as TBR items
(e.g., Geiselman et al., 1983), the FR participants in the present
experiment recognized TBF items less well (M ⫽ .65, sum of
rows 1 and 2 for FR participants in Table 1) than RR participants
recognized the corresponding TBR items (M ⫽ .77, sum of rows 1
and 2 for the RR participants in Table 1). This result appears to
arise from the increased episodic forgetting by the FR participants.
The bottom panel of Table 1 presents the proportions of
famous–yes, nonfamous–yes, and famous–no responses made to
old famous names by FR and RR participants. As these were TBR
items for all participants, it was not expected that the joint responses of the FR and RR participants would differ, and this
expectation was confirmed by the outcomes of planned comparisons, F(1, 40) ⫽ 0.02, MSE ⫽ 11.56, p ⬎ .50; F(1, 40) ⫽ 0.10,
MSE ⫽ 26.04, p ⬎ .50; F(1, 40) ⫽ 2.14, MSE ⫽ 234.5, p ⫽ .15,
for famous–yes, nonfamous–yes, and famous–no, respectively.
Thus, with respect to joint responses as well, it is only with old
nonfamous names that the fame judgments and memory performance of FR participants differed from that of RR participants. As
in Experiment 1, the judgments of FR participants have been
influenced to a greater extent by the indirect effects of information
they were told to forget than the judgments of RR participants
were influenced by the indirect effects of information they were
instructed to remember.
General Discussion
In Experiment 1, and strikingly replicated in Experiment 2,
participants instructed to forget nonfamous names presented during the memory task made significantly more misattributions of
real-world fame to such names during the subsequent famejudgment task than did participants instructed to continue remembering those same names. Apparently, old nonfamous names
530
BJORK AND BJORK
evoked feelings of familiarity in the fame-judgment task, and for
participants instructed to forget them, recollection of the source of
this familiarity was impaired, resulting in an increased tendency to
attribute it to prior real-world fame. Furthermore, their increased
susceptibility to the misleading effects of such familiarity seems to
stem from two types of source-memory problems: (a) impaired list
discrimination (i.e., an inability to recollect in which part of the
earlier memory task the old nonfamous name had occurred) and
(b) episodic forgetting per se (i.e., an inability to access a memory
representation for the old nonfamous name being presented during
the experiment).
Although unintuitive, this pattern of behavior is consistent with
our original speculation that under certain circumstances, the indirect memory effects of intentionally forgotten information might
be greater than those of corresponding intentionally remembered
information. As we speculated, individuals apparently become
susceptible to such enhanced indirect effects when, owing to
inhibited retrieval access for the intentionally forgotten information, the subsequent indirect or data-driven consequences of such
information cannot be checked or mitigated by accurate recollection of their source. Furthermore, the present findings suggest that
there may be two ways in which people become vulnerable to such
indirect effects of intentionally forgotten information. One is to
reduce what might be called episodic discrimination: Individuals
may know or be able to recognize that the intentionally forgotten
information was presented earlier as part of a particular episode
but have lost the ability to discriminate exactly when or where in
that episode it occurred. The second way arises from something
more like actual episodic forgetting: When the intentionally forgotten information continues to seem familiar, yet one is unable to
recollect that it occurred at all as part of a specific episode in the
past. In either of these cases, however, the indirect or unconscious
influences of the intentionally forgotten information remain in
effect and unchecked by accurate source recollection. Thus, these
indirect influences can have—as evidenced by the present results—a greater effect on certain behaviors than would be the case
were that same information to be intentionally remembered.
The evidence for continuing, and even enhanced, indirect influences of intentionally forgotten or inhibited information on subsequent behaviors has potentially important practical implications
regarding unintended consequences of instructions to forget in
certain legal and social settings. One might, for example, expect
sleeper-type effects to occur, not only owing to the inability of
individuals to recollect that the source of some argument was
unreliable with delay but also if, for some reason, they had been
instructed to forget that information or had even instructed themselves to forget it. Although well intended, the consequences of
such instructions to forget may be to strengthen the indirect effects
of the unreliable information.
The present findings also have implications for certain practices
in legal settings, particularly the practice by judges of instructing
jurors to disregard inadmissible evidence. Research concerned
with this practice has shown that instructions to disregard inadmissible evidence are not, by and large, successful (see Golding &
Long, 1998). Again, ironically, the present findings suggest that
the effects of such instructions to disregard inadmissible evidence
may be worse than useless; rather, they might actually increase the
indirect or unconscious influences of such evidence on jurors’ later
verdicts or judgments of guilt or innocence. Thus, the impact of
inadmissible evidence might prove more contaminating given instructions to forget that testimony than it would have been without
such an instruction.
Finally, although here and elsewhere we have often lauded the
virtues of retrieval inhibition as a unique and adaptive mechanism
for avoiding interference and resolving competition in memory as
well as other levels of cognitive processing (e.g., E. L. Bjork,
Bjork, & Anderson, 1998; R. A. Bjork, 1989), the present findings
suggest a dark side to retrieval inhibition as well. To the extent that
intentionally forgotten information remains in memory but recollective access to the source of that information is impaired, individuals become susceptible to misattributions of the familiarity or
processing fluency of that information when it is reencountered.
Those misattributions, in turn, can color certain judgments, as in
the legal and social examples discussed above, and produce errors,
as in the increased frequency of false-fame judgments reported in
the present research, and do so in ways that escape awareness.
References
Basden, B. H., Basden, D. R., & Gargano, G. J. (1993). Directed forgetting
in implicit and explicit memory tests: A comparison of methods. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 19,
603– 616.
Bjork, E. L., & Bjork, R. A. (1988). On the adaptive aspects of retrieval
failure in autobiographical memory. In M. M. Gruneberg, P. E. Morris,
& R. N. Sykes (Eds.), Practical aspects of memory: Current research
and issues: Vol. 1. Memory in everyday life (pp. 283–288). London:
Wiley.
Bjork, E. L., & Bjork, R. A. (1996). Continuing influences of to-beforgotten information. Consciousness and Cognition, 5, 176 –196.
Bjork, E. L., Bjork, R. A., & Anderson, M. C. (1998). Varieties of
goal-directed forgetting. In J. M. Golding & C. M. MacLeod (Eds.),
Intentional forgetting: Interdisciplinary approaches (pp. 103–137). Hillsdale, NJ: Erlbaum.
Bjork, E. L., Bjork, R. A., & Glenberg, A. (1973, November). Reinstatement of interference owing to to-be-forgotten items. Paper presented at
the 14th Annual Meeting of the Psychonomic Society, St. Louis, MO.
Bjork, E. L., Bjork, R. A., & Kirkpatrick, H. A. (1990, November). Direct
and indirect measures of inhibition in directed forgetting. Paper presented at the 31st Annual Meeting of the Psychonomic Society, New
Orleans, LA.
Bjork, R. A. (1972). Theoretical implications of directed forgetting. In
A. W. Melton & E. Martin (Eds.), Coding processes in human memory.
Washington, DC: Winston.
Bjork, R. A. (1989). Retrieval inhibition as an adaptive mechanism in
human memory. In H. L. Roediger & F. I. M. Craik (Eds.), Varieties of
memory and consciousness: Essays in honor of Endel Tulving (pp.
309 –330). Hillsdale, NJ: Erlbaum.
Block, R. A. (1971). Effects of instructions to forget in short-term memory.
Journal of Experimental Psychology, 89, 1–9.
Elmes, F. J., Adams, C., & Roediger, H. L. (1970). Cued forgetting in
short-term memory: Response selection. Journal of Experimental Psychology, 86, 103–107.
Geiselman, R. E., & Bagheri, B. (1985). Repetition effects in directed
forgetting: Evidence for retrieval inhibition. Memory & Cognition, 13,
51– 62.
Geiselman, R. E., Bjork, R. A., & Fishman, D. (1983). Disrupted retrieval
in directed forgetting: A link with posthypnotic amnesia. Journal of
Experimental Psychology: General, 112, 58 –72.
RESIDUAL INFLUENCES OF “FORGOTTEN” INFORMATION
Golding, J. M., & Long, D. L. (1998). There’s more to intentional forgetting than directed forgetting: An integrative review. In J. M. Golding &
C. M. MacLeod (Eds.), Intentional forgetting: Interdisciplinary approaches (pp. 103–137). Hillsdale, NJ: Erlbaum.
Gross, A. E., Barresi, J., & Smith, E. E. (1970). Voluntary forgetting of a
shared memory load. Psychonomic Science, 20, 73–75.
Jacoby, L. L., Kelley, C., Brown, J., & Jasechko, J. (1989). Becoming
famous overnight: Limits on the ability to avoid unconscious influences
of the past. Journal of Personality and Social Psychology, 56, 326 –338.
MacLeod, C. M. (1998). Directed forgetting. In J. M. Golding & C. M.
531
MacLeod (Eds.), Intentional forgetting: Interdisciplinary approaches
(pp. 1–57). Hillsdale, NJ: Erlbaum.
Reed, H. (1970). Studies of the interference processes in short-term memory. Journal of Experimental Psychology, 84, 452– 457.
Received June 17, 2002
Revision received December 17, 2002
Accepted December 28, 2002 䡲